Exercise-related interventions for cardiopulmonary function in adolescent idiopathic scoliosis: a systematic review and network meta-analysis

Abstract

Objective

To compare the effects of exercise-related interventions on cardiopulmonary function in adolescents with adolescent idiopathic scoliosis (AIS).

Methods

We searched PubMed, Embase, other databases, and trial registries for randomized controlled trials that reported cardiopulmonary outcomes in adolescents with AIS. Standardized mean differences (SMDs) were used for forced vital capacity (FVC) and forced expiratory volume in one second (FEV1), and mean differences (MDs) were used for the 6-minute walk distance (6MWD). Direct and indirect evidence was combined using a random-effects network meta-analysis. Intervention rankings were estimated using the surface under the cumulative ranking curve (SUCRA). Risk of bias was assessed with the revised Cochrane Risk of Bias 2 tool, and certainty of evidence was evaluated using Confidence in Network Meta-Analysis. Sensitivity analyses tested the effects of node composition, bracing-related interventions, postoperative or perioperative studies, risk of bias, and alternative assumptions for change-score standard deviations.

Results

24 randomized controlled trials involving 1221 adolescents with AIS were included. Brace combined with exercise training showed the most favourable ranking probability for FVC (SMD 2.87, 95% CI 0.24 to 5.50; SUCRA 93.2; low certainty). Mind–body exercise showed the most favourable ranking probability for FEV1 (SMD 3.28, 95% CI 0.43 to 6.13; SUCRA 92.5; low certainty). For 6MWD, resistance training showed the most favourable ranking probability (MD 143.14 m, 95% CI 110.39 to 175.89; SUCRA 99.3; moderate certainty), followed by aerobic endurance training (MD 121.14 m, 95% CI 93.56 to 148.71; SUCRA 86.7; moderate certainty). Overall, the certainty of evidence ranged from low to moderate. Some findings were limited by heterogeneity, sparse networks, small sample sizes, or possible publication bias.

Conclusion

Exercise-related interventions may improve pulmonary function and walking capacity in adolescents with AIS. However, because most comparative evidence was of low to moderate certainty and some rankings were affected by sparse networks and small-study effects, these findings should be interpreted as exploratory rather than as definitive exercise-prescription recommendations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13102-026-01887-3.

Introduction

Adolescent idiopathic scoliosis (AIS) is a three-dimensional structural deformity of the spine that develops during puberty. It is typically defined as a standing coronal Cobb angle of at least 10°, together with vertebral rotation and changes in sagittal spinal alignment [1]. Beyond visible trunk asymmetry, shoulder and pelvic imbalance, and possible back pain, AIS may also affect thoracic shape, breathing mechanics, and long-term cardiopulmonary health [2]. Epidemiological studies suggest that the screening prevalence of suspected scoliosis in children and adolescents is approximately 5.80%, while the global prevalence of radiographically confirmed idiopathic scoliosis is about 1.65%. The risk is higher among girls and adolescents with low body mass index, prolonged sedentary behaviour, longer screen exposure, and insufficient outdoor activity [3, 4]. Without timely detection and management during growth, spinal curvature may continue to progress in some patients and may later be associated with chronic pain, impaired physical function, reduced quality of life, and cardiopulmonary complications [5, 6].

The cardiopulmonary effects of AIS are closely related to changes in thoracic structure caused by the spinal deformity. Coronal curvature, vertebral rotation, and reduced thoracic kyphosis can alter rib alignment, thoracic volume, and chest wall mobility. These changes may restrict lung expansion and lead to a restrictive pattern of breathing impairment [7, 8]. Pulmonary function tests in patients with AIS often show reduced forced vital capacity (FVC), forced expiratory volume in one second (FEV1), and total lung capacity, while the FEV1/FVC ratio is usually relatively preserved. This suggests that impaired ventilation is mainly caused by restriction of the thoracic cage rather than airway obstruction [9, 10]. Thoracic deformity may also affect the diaphragm and intercostal muscles, reduce inspiratory and expiratory muscle strength, increase the work of breathing, and contribute to breathlessness and fatigue during exercise [11]. Cardiopulmonary exercise testing has further shown that patients with AIS may have reduced peak oxygen uptake and exercise tolerance. These impairments cannot be fully explained by Cobb angle alone, indicating that reduced muscle function, cardiovascular deconditioning, and low physical activity may also contribute to cardiopulmonary dysfunction [4, 12, 13].

Conservative treatment for AIS is usually guided by skeletal maturity, Cobb angle, and risk of curve progression. Common approaches include observation, bracing, and physiotherapeutic scoliosis-specific exercises (PSSE) [14]. Bracing can reduce curve progression and the risk of surgery, but long-term external restriction of the thorax may limit chest wall movement and affect breathing efficiency and respiratory muscle function [15]. For this reason, current management increasingly emphasizes not only control of spinal curvature, but also respiratory function, thoracic mobility, core stability, and overall exercise capacity [16]. Several exercise-based approaches, including Schroth, SEAS, Dobomed, inspiratory muscle training, aerobic training, and core stability training, have been used to improve spinal alignment, thoracic expansion, respiratory muscle strength, and exercise tolerance in patients with AIS [17–19]. These interventions, however, may work through different mechanisms. PSSE focuses on active three-dimensional correction and directed breathing, inspiratory muscle training aims to strengthen the breathing muscles, and aerobic training targets cardiovascular fitness and oxygen use. Understanding which intervention is most effective for each cardiopulmonary outcome is therefore important for developing individualized rehabilitation programmes.

Although previous systematic reviews suggest that exercise therapy may improve Cobb angle, trunk rotation, quality of life, and some pulmonary function measures in AIS, important gaps remain. Most reviews have focused mainly on spinal structure, while cardiopulmonary outcomes have often been treated as secondary findings and reported inconsistently [20, 21]. As a result, there is still no clear comparative framework for outcomes such as FVC, FEV1, respiratory muscle strength, peak expiratory flow, 6-minute walk distance, and peak oxygen uptake. Conventional pairwise meta-analyses are limited to interventions that have been directly compared within trials. Network meta-analysis is particularly suitable for the present research question because it allows multiple exercise-related intervention categories to be compared simultaneously by integrating both direct and indirect evidence, especially when direct head-to-head trials are limited [22]. Differences across studies in intervention type, training frequency, session duration, total treatment period, bracing status, and outcome measurement further complicate interpretation [23]. Therefore, this study used network meta-analysis to compare the effects of different exercise-related interventions on cardiopulmonary function in adolescents with AIS and to estimate their relative rankings across outcomes. The findings may provide a comparative evidence base for cardiopulmonary rehabilitation and individualized exercise prescription in adolescents with AIS.

Methods

Protocol and reporting guidelines

The protocol was developed before study screening and data extraction and was registered in PROSPERO (registration number: CRD420261390128). The study was designed, conducted, and reported in accordance with the PRISMA 2020 statement and the PRISMA extension for network meta-analysis. Full search strategies, study characteristics, risk-of-bias assessments, network diagnostics, CINeMA evaluations, and sensitivity analyses are provided in the Supplementary Materials.

Eligibility criteria

Eligibility criteria were defined using the PICOS framework. Participants were adolescents with idiopathic scoliosis, generally aged 10 to 19 years, diagnosed by clinical or radiographic assessment. Studies were also considered eligible when the participants were described by the original authors as having adolescent idiopathic scoliosis, even if the exact age range was not explicitly reported. No restrictions were applied regarding sex, country, ethnicity, or curve location. Eligible studies were randomized controlled trials that reported at least one cardiopulmonary outcome.

Interventions of interest were exercise-related treatments, including exercise alone and combined interventions in which exercise was a core component. These included physiotherapeutic scoliosis-specific exercises, respiratory muscle training, core stabilization training, aerobic endurance training, resistance training, proprioceptive neuromuscular facilitation, mind–body exercise, bracing combined with exercise, and auxiliary therapy combined with exercise. Eligible studies were required to include at least one exercise-related intervention arm. Comparators could include non-exercise or minimal-intervention controls, conventional exercise, bracing, manual or external therapy, or another exercise-related intervention. Non-exercise interventions such as bracing or manual therapy were retained only when they served as comparators within otherwise eligible randomized trials.

Studies were excluded if they were not randomized controlled trials, did not include adolescents with AIS, included predominantly adult or non-idiopathic scoliosis populations, did not report cardiopulmonary outcomes, or lacked sufficient data to extract or calculate means, standard deviations, and sample sizes. Reviews, conference abstracts, case reports, animal studies, non-clinical studies, duplicate publications, and repeated reports from the same study were also excluded. When multiple reports described the same study, the report with the most complete data or the most relevant follow-up period was retained.

Information sources and search strategy

A systematic search was conducted in PubMed, Embase, Web of Science, the Cochrane Library, CNKI, VIP, Wanfang, and SinoMed from database inception to April 30, 2026. The trial registries searched included ClinicalTrials.gov. Search terms were built around three concepts: AIS, exercise or rehabilitation interventions, and cardiopulmonary outcomes. The complete strategies are shown in Supplementary Appendix 1.

Study selection

All records were imported into reference management software, and duplicates were removed. Two reviewers independently screened titles and abstracts, followed by full-text assessment against the predefined eligibility criteria. Disagreements were resolved through discussion or, when necessary, by a third reviewer. Key characteristics of the included studies, including country, sample size, age, baseline Cobb angle, intervention node, intervention duration, and outcomes, are summarized in Supplementary Table S2.

Data extraction and treatment node classification

Two reviewers independently extracted data and cross-checked all entries. Extracted information included author, publication year, country, study design, sample size, age, baseline Cobb angle, intervention and comparator details, training frequency, intervention duration, follow-up time, outcomes, baseline and endpoint values, change scores, standard deviations, dropouts, adverse events, and risk-of-bias information.

To improve clinical interpretability and maintain network connectivity, intervention arms were grouped into treatment nodes according to their main therapeutic components and intended mechanisms. The final nodes were AE, AUX_EX, BR, BR_EX, CE, CS, CT, MB, MT, PNF, PSSE, RMT, and RT. CT referred to non-exercise or minimal-intervention control, CE to conventional exercise rehabilitation, PSSE to scoliosis-specific exercise based on active three-dimensional correction and curve-specific training, and RMT to training aimed primarily at improving respiratory muscle function or breathing mechanics. The treatment nodes and corresponding outcomes are presented in Supplementary Table S2, and the full data extraction items are listed in Supplementary Table S3.

Treatment nodes were finalized after intervention data extraction but before quantitative synthesis, based on the dominant therapeutic component and intended mechanism. When an intervention could fit more than one node, classification was determined by the component that distinguished it from the comparator. Thus, bracing combined with exercise was classified as BR_EX, exercise combined with a distinct non-exercise adjunct as AUX_EX, and multimodal exercise as the node representing its distinguishing active component. Differences in delivery format were not treated as separate nodes. Detailed classification rules and arm-level rationales are provided in Supplementary Table S2.2.

Outcomes

The primary outcomes were pulmonary ventilation measures, including FVC and FEV1. The key secondary outcome was the 6MWD, which reflects functional exercise capacity and cardiopulmonary endurance. Additional outcomes included maximal inspiratory pressure, maximal expiratory pressure, peak expiratory flow, maximal voluntary ventilation, and maximal oxygen uptake.

Because few studies reported maximal inspiratory pressure, maximal expiratory pressure, or peak expiratory flow, and because these outcomes did not form connected networks, they were not included in the network meta-analysis. Instead, they were summarized descriptively, as shown in Supplementary Appendix 4.

Risk-of-bias assessment

Risk of bias was assessed using the Cochrane Risk of Bias 2 tool. Five domains were evaluated: randomization, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. Each domain and the overall risk of bias were rated as low risk, some concerns, or high risk. Two reviewers completed the assessment independently, with disagreements resolved by discussion or adjudication by a third reviewer. The results are shown in Supplementary Table S5.

Statistical analysis

All statistical analyses were performed using Stata version 18.0 (StataCorp, College Station, TX, USA). Network meta-analyses, treatment rankings, league tables, inconsistency tests, comparison-adjusted funnel plots, and Egger’s tests were conducted within a frequentist framework using the Stata network meta-analysis suite. The certainty of evidence was assessed using the CINeMA web application.

Effect measures and data preparation

All outcomes were continuous variables. The primary analysis used change scores whenever possible. If change scores were not reported, they were calculated by subtracting baseline values from endpoint values. If the standard deviation of the change score was unavailable, it was estimated from baseline and endpoint standard deviations, assuming a correlation coefficient of 0.5. For outcomes reported as negative values, such as maximal inspiratory pressure, absolute values were used to ensure a consistent direction of effect.

Standardized mean differences were used for FVC and FEV1 because these outcomes were reported using different scales, including absolute values and percent-predicted values. For 6MWD, mean differences in metres were used because this outcome was consistently reported on the same metric scale and is directly clinically interpretable. Positive values indicated improvement.

Network meta-analysis model

For outcomes with connected networks, including FVC, FEV1, and 6MWD, random-effects network meta-analyses were performed to combine direct and indirect evidence. Network plots were used to show the available comparisons. For multi-arm trials, all eligible arms were included in a single network model; shared comparator groups were not duplicated, and correlations between effect estimates were accounted for within the frequentist NMA framework.

Analyses were conducted within a frequentist framework. Results were reported as relative effects with 95% confidence intervals. League tables were used to present all pairwise comparisons between interventions. The main league tables and complete effect estimates are provided in Supplementary Appendix 6.

Transitivity, heterogeneity, inconsistency, and small-study effects

Before conducting the network meta-analysis, we assessed transitivity by comparing the distributions of age, baseline Cobb angle, intervention duration, bracing status, and perioperative status across treatment nodes using the available study-level data. The distributions of age and baseline Cobb angle, summarized in Supplementary Table S7.1, showed no clear systematic differences between nodes. However, several potential effect modifiers were incompletely reported or unevenly distributed. Subgroup analysis and meta-regression were not feasible because most nodes included few studies, and further stratification would have produced sparse or disconnected networks. Transitivity was therefore considered plausible, although it could not be fully verified.

Heterogeneity was assessed using τ² from the random-effects model. Global inconsistency was examined using a global inconsistency test, and local inconsistency was assessed using node-splitting or side-splitting methods. Small-study effects and potential publication bias were evaluated using comparison-adjusted funnel plots and Egger’s test. The results for heterogeneity, inconsistency, node-splitting, and Egger’s test are presented in Supplementary Appendices 8 and 9.

Treatment ranking

Interventions were ranked using the surface under the cumulative ranking curve (SUCRA). A higher SUCRA value indicates a greater probability that an intervention is among the most effective options for a given outcome. Ranking results and cumulative probability plots are shown in Supplementary Appendix 10.

Certainty of evidence, sensitivity analyses, and descriptive synthesis

The certainty of evidence was assessed using the CINeMA framework across six domains: within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and inconsistency. Certainty for each comparison was rated as high, moderate, low, or very low. CINeMA results for FVC, FEV1, and 6MWD are provided in Supplementary Appendix 7.

Sensitivity analyses were conducted by excluding combined or adjunctive treatment nodes, mind–body exercise, bracing-related nodes, postoperative or perioperative studies, and studies at high risk of bias. Additional analyses tested alternative assumed correlation coefficients for change-score SDs (r = 0.25 and r = 0.75). Low-risk-only and small-study-exclusion analyses were attempted, but overall network estimates were not pooled when the resulting networks became too sparse or disconnected. Their definitions and results are provided in Supplementary Appendix 11.

For outcomes that did not form connected networks, including maximal inspiratory pressure, maximal expiratory pressure, and peak expiratory flow, network meta-analysis and treatment ranking were not performed. These outcomes were summarized descriptively by direct comparison, and the direction of effect was used only to describe trends rather than statistical significance. These results are presented in Supplementary Appendix 4.

Results

Study selection and characteristics of included studies

The database and trial registry searches identified 3900 records, including 3867 from databases and 33 from trial registries. After removing 1240 duplicates, 2660 records were screened by title and abstract. Of these, 2578 were excluded, leaving 82 full-text articles for eligibility assessment. A further 58 articles were excluded because they were not randomized controlled trials (n = 23), did not include patients with AIS (n = 15), did not report cardiopulmonary outcomes (n = 11), provided insufficient data (n = 5), or were not original clinical studies (n = 4). Finally, 24 randomized controlled trials involving 1221 patients with AIS were included.

The studies evaluated several exercise-related intervention nodes, including PSSE, RMT, CS, AE, RT, PNF, MB, AUX_EX, BR_EX, and control nodes such as CT or CE. The study selection process is shown in Fig. 1, and detailed study characteristics are provided in Supplementary Table S2.

Fig. 1.

PRISMA flow diagram

Risk-of-bias assessment

Overall, most included trials were rated as having either low risk of bias or some concerns. Seven studies were judged to have low overall risk of bias, 15 had some concerns, and 2 had high risk of bias. The RoB 2 results are shown in Fig. 2, and the full assessment is provided in Supplementary Table S5.

Fig. 2.

RoB 2 risk-of-bias assessment: (A) summary plot; (B) traffic-light plot

Transitivity, heterogeneity, inconsistency, and small-study effects

The transitivity assumption was assessed by comparing key clinical and methodological characteristics across intervention nodes, including age, baseline Cobb angle, disease stage, bracing status, postoperative or perioperative status, and intervention duration. Mean age was broadly similar across most nodes, with an overall weighted mean of 14.39 ± 2.56 years. The overall weighted mean baseline Cobb angle was 31.13 ± 18.29°. However, some differences were observed. For example, the AE and RT nodes had relatively higher baseline Cobb angles, while participants in the MB node were slightly older. The distributions of age and baseline Cobb angle are shown in Supplementary Table S7.1.

Model diagnostics showed that FVC, FEV1, and 6MWD all formed connected networks. Global inconsistency tests showed P values of 0.11, 0.20, and 0.09 for FVC, FEV1, and 6MWD, respectively, with corresponding τ² values of 1.116, 1.311, and < 0.001. Local inconsistency was further assessed using node-splitting. Some direct and indirect comparisons showed potential disagreement, indicating that several network estimates should be interpreted cautiously. Egger’s tests suggested possible small-study effects for FVC and FEV1, with P values of 0.04 and 0.003, respectively, but not for 6MWD after reanalysis using mean differences (P = 0.164). Full results for heterogeneity, inconsistency, node-splitting, and funnel plots are provided in Supplementary Appendices 8 and 9. These imbalances suggest that indirect comparisons involving these nodes should be interpreted cautiously.

Network meta-analysis results

Network structures and forest plots versus CT are shown in Figs. 3, 4 and 5. Full league tables and CINeMA assessments are provided in Supplementary Appendices 6 and 7, respectively; SUCRA rankings and cumulative probability plots are provided in Supplementary Appendix 10.

Fig. 3.

Network meta-analysis for FVC: (A) network plot; (B) forest plot versus CT

Fig. 4.

Network meta-analysis for FEV1: (A) network plot; (B) forest plot versus CT

Fig. 5.

Network meta-analysis for 6MWD: (A) network plot; (B) forest plot versus CT

FVC

The FVC network included 20 direct comparisons and 11 intervention nodes. The network plot showed that PSSE, CE, and AUX_EX formed the main comparison framework.

Compared with CT, point estimates for most interventions favoured improvement in FVC, although several confidence intervals crossed the null value. BR_EX had the largest effect estimate for FVC, but the estimate was based on indirect evidence and had a wide confidence interval (SMD 2.87, 95% CI 0.24 to 5.50; SUCRA 93.2; Low certainty). MB and AUX_EX also showed favourable directions of effect, but their confidence intervals were wide.

CINeMA indicated that the certainty of evidence for FVC was mainly low to moderate. These findings should be interpreted alongside the confidence intervals and sensitivity analyses.

FEV1

The FEV1 network included 22 direct comparisons and 12 intervention nodes. PSSE, AUX_EX, and CE were the main core nodes and were connected with RMT, CS, RT, MB, CT, and other interventions through direct or indirect comparisons. This network provided a basis for comparing the relative effects of different exercise-related interventions on FEV1.

Compared with CT, point estimates for most interventions favoured improvement in FEV1, although several confidence intervals crossed the null value. MB had the largest effect estimate, based on a single direct comparison, although the confidence interval was also wide (SMD 3.28, 95% CI 0.43 to 6.13; SUCRA 92.5; Low certainty). AUX_EX and RT also showed favourable directions of effect, but their confidence intervals were wide (AUX_EX: SMD 2.03, 95% CI − 0.09 to 4.15; SUCRA 77.7; Moderate certainty; RT: SMD 1.79, 95% CI − 1.59 to 5.18; SUCRA 65.7; Low certainty). CS and PSSE also showed positive point estimates compared with CT.

CINeMA indicated that the certainty of evidence for FEV1 was mainly low to moderate. The findings should be interpreted cautiously because of the limited number of studies, wide confidence intervals, and sensitivity analysis findings.

6MWD

The 6MWD network included 11 direct comparisons and 10 intervention nodes. AUX_EX and PSSE were the main nodes with larger sample contributions, while AE, RT, RMT, MT, PNF, CS, CE, and CT formed the remaining comparison network for functional exercise capacity.

Compared with CT, point estimates for several interventions favoured improvement in 6MWD when analysed using mean differences in metres, although some confidence intervals crossed the null value. RT had the largest effect estimate, supported by indirect evidence through AE (MD 143.14 m, 95% CI 110.39 to 175.89; SUCRA 99.3; Moderate certainty), followed by AE (MD 121.14 m, 95% CI 93.56 to 148.71; SUCRA 86.7; Moderate certainty). AUX_EX, PSSE, and RMT also showed favourable directions of effect, with confidence intervals that did not include zero (AUX_EX: MD 85.73 m, 95% CI 28.60 to 142.86; SUCRA 78.7; Low certainty; PSSE: MD 62.86 m, 95% CI 6.21 to 119.52; SUCRA 61.6; Low certainty; RMT: MD 59.77 m, 95% CI 17.46 to 102.08; SUCRA 57.7; Low certainty).

CINeMA showed that the certainty of evidence for 6MWD ranged from low to moderate for most comparisons. Although the use of mean differences improved clinical interpretability, the findings should be interpreted cautiously because several comparisons were informed by sparse evidence.

Sensitivity analyses and descriptive synthesis of additional outcomes

Sensitivity analyses showed that some rankings were affected by node composition and change-score SD assumptions. After excluding combined or adjunctive nodes, PSSE ranked highest for FVC and FEV1, whereas RT ranked highest for 6MWD. Analyses using r = 0.25 and r = 0.75 changed the magnitude of some estimates but did not materially alter the overall interpretation. Analyses restricted to nonoperative AIS populations showed that the main findings were generally similar, although rankings for some outcomes changed. Low-risk-only and small-study-exclusion analyses could not be pooled because the networks were too sparse or disconnected. Full definitions and results of the sensitivity analyses are provided in Supplementary Appendix 11.

MIP, MEP, and PEF were summarized descriptively because they did not form connected networks. RMT, RT, and CS showed favourable numerical changes in some studies, whereas PEF results were less consistent. These outcomes were considered supportive only and were not used for treatment ranking. Full descriptive synthesis results are provided in Supplementary Appendix 4.

Discussion

Main findings

This network meta-analysis combined direct and indirect evidence from randomized controlled trials to compare exercise-related interventions for cardiopulmonary outcomes in adolescents with AIS. While previous reviews have mainly focused on structural outcomes, such as Cobb angle and trunk rotation, or on health-related quality of life, this study focused on FVC, FEV1, and 6MWD. These outcomes provide complementary information on pulmonary function and functional exercise capacity, both of which are important in comprehensive AIS management.

Overall, exercise-related interventions tended to favour improvements in cardiopulmonary outcomes, but the most favourable intervention differed by outcome. BR_EX ranked highest for FVC, MB ranked highest for FEV1, and RT ranked highest for 6MWD. However, these rankings should be interpreted as exploratory because many comparisons were supported by low- or moderate-certainty evidence. Sensitivity analyses showed that BR_EX remained favourable for FVC in most analyses, MB remained favourable for FEV1 in several analyses, and RT remained favourable for 6MWD, but rankings changed after excluding combined or adjunctive nodes and under alternative change-score SD assumptions. Low-risk-only and small-study-exclusion analyses could not be pooled because the networks became too sparse or disconnected.

Comparison with previous studies

The present findings extend previous systematic reviews and meta-analyses of conservative treatment for AIS. Most existing reviews have focused on spinal deformity or patient-reported outcomes, whereas cardiopulmonary outcomes have received less systematic attention [20, 21, 23]. By assessing FVC, FEV1, and 6MWD within a connected treatment network, this study provides a more focused comparison of rehabilitation strategies for cardiopulmonary function in AIS.

The results also suggest that interventions effective for structural correction may not necessarily be the most effective for improving lung function or walking capacity. Although PSSE remain central to AIS rehabilitation, they did not consistently rank first across cardiopulmonary outcomes in this analysis. Instead, BR_EX ranked highest for FVC, MB ranked highest for FEV1, and RT ranked highest for 6MWD. These findings indicate that structural correction and physiological rehabilitation are related but distinct treatment goals.

Our findings are consistent with recent pairwise meta-analyses showing that exercise therapy may improve pulmonary function in patients with mild to moderate AIS. Chen, Xie, and colleagues reported that short-term exercise interventions improved FEV1, FVC, maximal inspiratory pressure, and maximal expiratory pressure in patients with AIS [24]. Similarly, Zhao et al. found that exercise combined with other conservative therapies improved pulmonary function and spinal alignment in the short term [25]. Compared with pairwise meta-analyses, the present network meta-analysis adds further information by comparing multiple intervention categories and identifying outcome-specific patterns.

Earlier reviews of the Schroth method reported inconsistent effects on FVC, with some pooled estimates not reaching statistical significance [26]. This may be explained by differences in intervention classification, sample size, and analytical approach. Pairwise analyses limited to a single exercise approach may have limited power and may not capture the potential benefits of combined or multimodal interventions. In contrast, this study included several intervention categories, including BR_EX, MB, RT, and AUX_EX, allowing a broader comparison of rehabilitation strategies. These findings suggest possible outcome-specific effects, but they require confirmation in larger, high-quality randomized trials.

Potential mechanisms

Cardiopulmonary impairment in AIS is commonly linked to three-dimensional spinal deformity, altered thoracic mechanics, reduced chest wall mobility, and impaired respiratory muscle function. In a restrictive breathing pattern, FVC and FEV1 may decrease while the FEV1/FVC ratio remains relatively preserved, suggesting that limited thoracic expansion, rather than airway obstruction, is the main problem. Cardiopulmonary exercise testing has also shown that adolescents with AIS may have lower peak oxygen uptake and minute ventilation than healthy peers, and that these impairments cannot be explained by Cobb angle alone [12]. Reduced muscle function, impaired postural control, and lower physical activity may also contribute to cardiopulmonary dysfunction.

The favourable effect of BR_EX on FVC may reflect the combined effect of external support and active three-dimensional breathing training. Although rigid bracing alone may restrict chest wall movement and reduce lung volume in some patients, modern three-dimensional braces, such as the Chêneau brace, are designed to apply corrective pressure while preserving expansion space on the concave side of the thorax [27]. When combined with scoliosis-specific breathing techniques, such as rotational angular breathing, the brace may serve not only as a corrective device but also as a guide for directed thoracic expansion. Therefore, BR_EX should be understood as a brace-based combined intervention rather than as evidence for exercise alone. Whether structured breathing or exercise training should be routinely integrated into brace management needs confirmation in larger trials.

The high ranking of MB for FEV1 may be related to its focus on controlled breathing, core stability, postural awareness, and movement coordination. AIS is increasingly understood not only as a structural spinal deformity but also as a condition involving altered postural control, proprioception, and sensorimotor integration [28]. Pilates- and yoga-based interventions often require coordinated breathing, activation of deep trunk muscles, and sustained attention to body alignment [29, 30]. These features may improve respiratory muscle coordination, including the diaphragm, intercostal muscles, abdominal wall, and deep spinal stabilizers. This may partly explain the favourable ranking of MB for FEV1, although the evidence remains limited and heterogeneous.

The improvements in 6MWD observed with RT, aerobic endurance training (AE), and AUX_EX are likely related to broader functional gains rather than changes in pulmonary mechanics alone. The 6MWD reflects overall functional capacity, including cardiopulmonary reserve, lower-limb strength, trunk control, gait efficiency, and resistance to fatigue. Resistance training may improve muscle strength, oxidative capacity, and movement efficiency, thereby reducing the effort required for walking and daily activities [31, 32]. AUX_EX may also improve walking capacity by reducing pain, increasing mobility, and improving gait mechanics [33]. These effects may enhance functional performance even when changes in resting pulmonary function are modest.

Strengths and limitations

This study has several strengths. First, it is the first network meta-analysis to compare exercise-related interventions specifically for cardiopulmonary outcomes in AIS, rather than focusing mainly on spinal deformity or quality of life. Second, it included both pulmonary and functional outcomes and clearly distinguished the main analyses of FVC, FEV1, and 6MWD from the descriptive synthesis of additional outcomes. Third, the network approach allowed multiple interventions to be compared within a single analytical framework by combining direct and indirect evidence. Fourth, the certainty of evidence was assessed using CINeMA, and intervention classification was supported by an intervention-component matrix. Finally, several sensitivity analyses were conducted to examine the influence of node composition, postoperative or perioperative studies, risk of bias, and alternative assumptions for change-score standard deviations.

Several limitations should also be considered. First, although connected networks were available for FVC, FEV1, and 6MWD, the evidence was insufficient for robust network analysis of other important outcomes, such as maximal inspiratory pressure, maximal expiratory pressure, and peak expiratory flow. These outcomes are closely related to respiratory muscle strength and dynamic airway function. Their limited availability restricts a fuller understanding of the mechanisms underlying pulmonary improvement. Future trials should include standardized assessments of MIP, MEP, PEF, and cardiopulmonary exercise testing outcomes, such as peak oxygen uptake and ventilatory efficiency [12, 34].

Second, differences in intervention protocols and outcome measurement may have affected the pooled estimates. Studies varied in exercise content, supervision, intensity, treatment duration, brace use, and outcome reporting. Between-study heterogeneity was substantial for FVC and FEV1 but minimal for 6MWD, possibly because 6MWD was analysed using mean differences in metres. We did not identify established AIS-specific minimal clinically important differences for FVC, FEV1, or 6MWD. Therefore, statistically significant results, especially large standardized effects for FVC and FEV1, should not be interpreted automatically as clinically important changes.

Third, the overall certainty of evidence was limited by small sample sizes, possible publication bias, methodological challenges common in exercise-based trials, and relatively short follow-up periods. Many included studies were small, single-centre randomized controlled trials, and blinding of participants and therapists is difficult in physical therapy research. Funnel plot asymmetry and Egger’s tests suggested possible small-study effects for some outcomes. In addition, the durability of cardiopulmonary benefits remains uncertain because most studies had short follow-up periods. Therefore, although the rankings identify promising interventions, they should be interpreted cautiously. Future larger trials should evaluate both intervention type and dose, while dose–response network meta-analysis may clarify how intensity, frequency, session duration, total intervention period, and weekly exercise volume affect different outcomes [35, 36].

Conclusion

Exercise-related interventions may improve pulmonary function and functional exercise capacity in adolescents with AIS. However, most comparative evidence was of low to moderate certainty, and several rankings were influenced by sparse networks, heterogeneity, wide confidence intervals, or possible small-study effects. These findings identify promising intervention categories rather than definitive clinical prescriptions and require confirmation in larger, well-designed AIS trials evaluating both intervention type and exercise dose.

Abbreviations

AE

Aerobic endurance training

AIS

Adolescent idiopathic scoliosis

AUX_EX

Auxiliary therapy combined with exercise

BR

Bracing

BR_EX

Bracing combined with exercise

CE

Conventional exercise

CINeMA

Confidence in Network Meta-Analysis

CS

Core stabilization training

CT

Non-exercise or minimal-intervention control

FEV1

Forced expiratory volume in 1 s

FVC

Forced vital capacity

MB

Mind-body exercise

MT

Manual or external therapy

NMA

Network meta-analysis

PNF

Proprioceptive neuromuscular facilitation

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PSSE

Physiotherapeutic scoliosis-specific exercises

RMT

Respiratory muscle training

RoB 2

Revised Cochrane risk-of-bias tool for randomized trials

RT

Resistance training

SMD

Standardized mean difference

MD

Mean difference

SUCRA

Surface under the cumulative ranking curve

6MWD

6-minute walk distance

Authors’ contributions

Conceptualization: Xiaoquan Weng, Wenrui Huang, Lingyan Yuan. Methodology: Xiaoquan Weng, Wenrui Huang, Lingyan Yuan. Data curation: Xiaoquan Weng, Shuo Zhou. Formal analysis: Xiaoquan Weng. Investigation: Xiaoquan Weng, Shuo Zhou. Writing – original draft: Xiaoquan Weng. Writing – review & editing: Xiaoquan Weng, Shuo Zhou, Wenrui Huang, Lingyan Yuan. Supervision: Lingyan Yuan, Wenrui Huang.

Funding

This work was supported by the Humanities and Social Sciences Planning Fund of the Ministry of Education under the project “Dose–Response Relationship Between Physical Activity and Spinal Health in Children and Adolescents” (24YJAZH207) and the Shanghai Philosophy and Social Sciences Planning Project under the project “Risk Prediction and Multi-Stakeholder Collaborative Intervention for Abnormal Spinal Curvature in Adolescents.” (2022BTY003). The funders provided financial support only and had no role in the study design, data collection, data analysis, data interpretation, or preparation of the manuscript.

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files. Additional data related to the analyses are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.